A web page is a document
once it is text.
Droblet Intelligence runs the dScribe engine on the web. Give it a URL and a schema and it returns validated JSON. Give it a description of your ideal customer and it searches, crawls company sites, extracts profiles and contact details, scores the fit and hands qualified leads to the sales agent.
URL + schema
In, validated JSON out
8
Shipped web schemas, plus any of yours
Per unit
Failure: one bad site never stops a campaign
One engine, two products
The hard part of reading a web page is the same as the hard part of reading a PDF: turning text into fields you can trust. Droblet Intelligence reuses the dScribe engine for that, so schemas, confidence scores, business rules and the review loop all apply unchanged. The only new work is fetching the page and turning it into text.
Web Intelligence is the direct form: a URL and a schema in, JSON out. It crawls the pages that matter on a site (home, about, contact, pricing, services), extracts the fields, and validates them. A blocked or empty fetch escalates to a rendering proxy by itself.
Lead Intelligence is the campaign form. Describe what you sell and who buys it; it searches, finds company websites, builds a profile of each, pulls emails, phones and social links from the page itself (parsed from the HTML, not guessed by the model), scores each company against your ideal customer with a written reason, and hands the ones above your threshold to Droblet Agent to work.
What it does
Web extraction
Any page, any schema. Validated JSON by webhook or pull.
Site crawl
Home plus the about, contact, pricing and services pages, as readable text with the signals (emails, phones, socials) kept separately.
Deterministic first
Contact details are parsed from the HTML and fed to the model as facts. For leads, parsed facts win over model output.
Lead campaigns
Search, crawl, profile, enrich, qualify. Each prospect carries a fit score and the reasoning behind it.
Fails per unit
A site that times out marks its own extraction failed. The campaign carries on.
Keyless in development
Direct fetch and DuckDuckGo keep development free of API keys; a rendering proxy and SERP API take over in production.
Common questions
Droblet fetches public pages, honours robots directives and rate limits, and identifies itself. You are responsible for how you use what it finds; we build in the controls to use it carefully.
Contact details are taken from the page, not invented. The fit score is the model's judgement with its reasoning written down, so a person can disagree with it in one glance.
No. Intelligence runs on the dScribe engine and can be licensed on its own.
Run one campaign with us
Describe your ideal customer. We will show you the first twenty qualified companies and why they qualified.